Cardiovascular diseases are the leading cause of death globally. A significant number of these diseases are treated with medical devices that are in contact with blood. The use of artificial biomaterials promotes the formation of blood clots (thrombosis) which are the most detrimental adverse events related to medical devices. Furthermore, these systems perturbate the physiological fluid dynamics stimulating biochemical pathways that contribute to the thrombogenic process. Mathematical models are used to predict the regions prone to thrombosis with the short-term goal of optimizing devices and in the long term inform patient specific treatments. This is, however, a very ambitious and challenging task, as the thrombogenic process involves hundreds of biochemical reactions that take place in a complex fluid (blood) involving a wide variety of spatial and temporal scales. In this context, data driven methods such as reduced order modeling, data assimilation, and uncertainty quantification are necessary to perform high fidelity multi-scale simulations in contemporary medical devices. The current talk will present thrombosis modeling aspects including fluid mechanics and biochemistry, a clotting assay simulation, an uncertainty quantification study, and preliminary results on reduced order modeling combining proper orthogonal decomposition and a neural network architecture.